REVERSE ENGINEERING THE YOUTUBE VIDEO DELIVERY CLOUD. Vijay Kumar Adhikari, Sourabh Jain, Yingying Chen, and Zhi-Li Zhang

Size: px
Start display at page:

Download "REVERSE ENGINEERING THE YOUTUBE VIDEO DELIVERY CLOUD. Vijay Kumar Adhikari, Sourabh Jain, Yingying Chen, and Zhi-Li Zhang"

Transcription

1 REVERSE ENGINEERING THE YOUTUBE VIDEO DELIVERY CLOUD Vijay Kumar Adhikari, Sourabh Jain, Yingying Chen, and Zhi-Li Zhang Department of Computer Science & Engineering, University of Minnesota - Twin Cities ABSTRACT In this paper we set out to reverse-engineer the YouTube video delivery cloud by building a globally distributed active measurement infrastructure. Through careful and extensive data collection, analysis and experiments, we deduce the key design features underlying the YouTube video delivery cloud. The design of the YouTube video delivery cloud consists of three major components: a flat video id space, multiple DNS namespaces reflecting a multi-layered logical organization of video servers, and a 3-tier physical cache hierarchy. By mapping the video id space to the logical servers via a fixed hashing and cleverly leveraging DNS and HTTP redirection mechanisms, such a design leads to a scalable, robust and flexible content distribution system. 1. INTRODUCTION Given the traffic volume, geographical span and scale of operations, the design of YouTube s content delivery infrastructure is perhaps one of the most challenging engineering tasks (in the context of most recent Internet development). Before Google took over YouTube in late 2006 [1] and subsequently re-structured the YouTube video delivery infrastructure, it was known that YouTube employed several data centers in US (see [3]) as well as third-party content delivery networks [2, 9] to stream videos to users. Since Google s takeover, YouTube has grown rapidly and became several-fold larger both in terms of users and videos. For instance, using inter-domain traffic collected in 2007 and 2009 at hundreds of ISPs across the world, the authors of a recent study [6] show that Google has become one of the top five inter-domain traffic contributors in 2009; a large portion of Google s traffic can be attributed to YouTube. While it is widely expected that Google has re-structured and incorporated the YouTube delivery system into its own vast Internet infrastructure in the past few years, little is known how Google leverages its resources to re-design and re-structure the YouTube video delivery infrastructure which we will refer to as the YouTube video delivery cloud to meet the rapidly growing user demands as well as user performance expectations. This work is supported in part by the NSF grants CNS and CNS , and the DTRA Grant HDTRA This paper attempts to reverse-engineer the YouTube video delivery cloud through large-scale active measurement, data collection and analysis. We are particularly interested in answering the following important design questions: i) how does YouTube design and deploy a scalable and distributed delivery infrastructure to match the geographical span of its users and meet varying user demands? ii) how does YouTube perform load-balancing across its large pool of Flash video servers (and multiple locations)? and iii) given the sheer volume of YouTube videos which renders it too costly, if not nearly impossible, to replicate content at all locations, what strategies does YouTube use to quickly find the right content to deliver to users? On one hand we believe that Google s YouTube video delivery cloud offers an example of the best practices in the design of an Internet-scale content delivery infrastructure. On the other hand, the design of YouTube video delivery cloud also poses some interesting and important questions regarding alternative architectural designs, cache placement, content replication and load balancing strategies, and so forth. The global scale of the YouTube video delivery cloud poses several challenges in actively measuring, and collecting data from the YouTube video delivery cloud. To address these challenges, we have developed a novel distributed active measurement platform with more than 1000 vantage points spanning five continents. Through careful data analysis and inference especially by analyzing the relations among YouTube video ids, DNS names, and IP addresses and by conducting extensive experiments to test and understand the behavior of the YouTube video delivery cloud, we are not only able to geo-locate a large portion of YouTube video server and cache locations, but also to uncover and deduce the logical designs of the YouTube video id space, the DNS namespace structures and cache hierarchy, and how they map to the physical infrastructure and locations. We describe the measurement infrastructure and collected data in Section 3. We provide a summary of the key findings regarding the YouTube design in Section 4. In Section 5 and Section 6 we present more details regarding how we derive these findings, including analysis performed, the methods used, and additional experiments conducted to verify and validate the findings.

2 2. RELATED WORK Most existing studies of YouTube mainly focus on user behaviors or the system performance. For instance, the authors in [4] examined the YouTube video popularity distribution, popularity evolution, and its related user behaviors and key elements that shape the popularity distribution using datadriven analysis. The authors in [5] investigate the YouTube video file characteristics and usage patterns such as the number of users, requests, as seen from the perspective of an edge network. Another study [9] analyzed network traces for YouTube traffic at a campus network to understand benefits of alternative content distribution strategies. A more recent work [8] studies the impact of the YouTube video recommendation on the popularity of videos. Perhaps most relevant to our work is the recent study carried in [3], where the authors utilize the Netflow traffic data passively collected at various locations within a tier-1 ISP to uncover the locations of YouTube data center locations, and infer the load-balancing strategy employed by YouTube at the time. The focus of the study is on the impact of YouTube load-balancing on the ISP traffic dynamics, from the perspective of the tier-1 ISP. As the data used in the study is from spring 2008, the results reflect the YouTube delivery infrastructure pre Google re-structuring. To the best of our knowledge, our work is the first study that attempts to reverse engineer the current YouTube video delivery cloud. Fig. 1. Our Active Measurement Platform: an Illustration. process including multiple HTTP request redirections and DNS resolutions. The detailed text-based logs recorded for each step of the process contain a variety of information such as the hostnames and URLs involved in each step, the HTTP request and response messages and their status codes, the basic HTML payload and timestamps for each of the steps. In addition, our emulated YouTube Flash video player can be configured to use an open recursive DNS servers (instead of the default local DNS server of a PlanetLab node). This capability therefore enables us to use the 843 open recursive DNS servers as additional vantage points. Hence we have a total of 1, 314 globally distributed vantage points for active YouTube measurement and data collection. 3. MEASUREMENT PLATFORM In this section we first briefly describe the basics of YouTube video delivery. We then provide an overview of our distributed active measurement and data collection platform YouTube Video Delivery Basics When a user visits, or clicks on any URL of the form the web server returns an HTML page with embedded URLs of certain forms, e.g., v1.lscache5.c.youtube.com, pointing to the Flash video server responsible for serving that video. When the user clicks the playback button of an embedded Flash video object on the page, the browser resolves the hostname to get an IP address for the Flash video server, which then streams the video to the user s browser Active Measurement Platform Our active measurement platform utilizes 471 PlanetLab nodes, and 843 open recursive DNS servers provided by and located at various ISPs and organizations. We also developed an emulated YouTube Flash video player in Python which emulates the process involved in playing back a YouTube video using HTTP. The emulator records detailed log of the 3.3. Measurement Methodology and Datasets We adopt a multi-step process to collect, measure, and analyze YouTube data. First, we crawl the YouTube website from geographically dispersed vantage points using the PlanetLab nodes to collect a list of videos, and record their view counts and other relevant metadata. We show the popularity of the videos in our list in Fig. 3. We note that because of our sampling method, a large percentage of the videos in our list are popular videos. To address this problem, we added a large number of videos with very low view counts to this list of videos. Second, we feed the URLs referencing the videos to our emulated YouTube Flash video players, download and playback the Flash video objects from the 471 globally distributed vantage points, perform DNS resolutions from these vantage points, and record the entire playback processes including HTTP logs. Third, we perform ping-based latency measurements from the PlanetLab nodes to all the observed IP addresses. Furthermore, we also extract the HTTP request redirection sequences, analyze and model these sequences to understand YouTube redirection logic. 4. SUMMARY OF KEY FINDINGS In this section we provide a summary of our key findings regarding the design and operations of the YouTube global

3 Fig. 2. Geographical distribution of YouTube Video Cache Locations. video delivery system. This serves as the road map for the ensuing sections, where we will provide specifics as to how we arrive at these findings, including the data analysis and inference as well as experiments we have performed to verify the findings Overall Architecture The design of YouTube video delivery cloud consists of three major components: the video id space, the multi-layered organization of logical video servers via multiple anycast DNS namespaces, and a 3-tier physical server cache hierarchy with (at least) 38 primary locations, 8 secondary locations and 5 tertiary locations. Here by a anycast (DNS) namespace we mean that each DNS name is by design, mapped to multiple IP addresses. YouTube Video Id Space. Each YouTube video is uniquely identified using a flat identifier of 11 literals long, where each literal can be [A-Z], [0-9], - or. The total size of the YouTube video id space is effectively Analyzing the 434K video ids we collected, we find that they are uniformly distributed in the video id space. Anycast DNS Namespaces and Layered Logical Video Server Organization. Based upon the request redirection mechanism deiscussed later, we observe that YouTube defines multiple (anycast) DNS namespaces, each representing a collection of logical video servers with certain roles. Together, these (anycast) DNS namespaces form a layered organization of logical video servers. As shown in Table 1, there are a total of three anycast namespaces, which we refer to as lscache, tccache, and cache namespaces; each namespace has a specific format 1. Physical Server Cache Hierarchy and Their Locations. Using the YouTube IP addresses seen in our datasets, we are able to geo-map the YouTube physical video server cache locations, which are dispersed at five continents (see Fig. 2). We next see what logical hostnames map to what location to classify the physical cache locations. From this analysis, 1 We also encountered, although very rarely, slightly different formats for lscache and cache namespaces where hostnames had nonxt and altcache in them respectively we deduce that YouTube employs a 3-tier physical cache hierarchy with (at least) 38 primary locations, 8 secondary locations and 5 tertiary locations. Each location contains varying number of IP addresses ( physical video servers), and there are some overlapping between the primary and secondary locations (e.g., at the Washington D.C. metro areas), where one physical video server may serve either as a primary or a secondary video server. Columns 4-6 show the total number of IPs, prefixes, and locations each DNS namespace is mapped. In Section 5.3 we will provide some details regarding how we geo-map the YouTube physical cache locations. Unicast Namespace. In addition, for each IP address, YouTube also defines a unicast DNS name. Namely, there is a one-to-one between this DNS name and the IP address. As shown in Table 1, the unicast names have two formats, which we refer to as rhost and rhostisp formats Mechanisms and Strategies The introduction of the layered organizations of logical video servers via multiple namespaces enables YouTube to employ several mechanisms and strategies to i) map videos to logical video servers via a fixed hashing, and ii) map logical video servers to physical video servers at various locations of its physical cache hierarchy through both DNS resolution and HTTP redirection mechanisms. Fixed Mapping between Video Id Space and Logical Video Servers (Anycast DNS Namespaces). YouTube adopts a fixed hashing to map each video id uniquely to one of the 192 DNS names in the lscache namespace. In other words, the video id space is uniformly divided into 192 sectors, and each lscache DNS name representing a logical video cache server is responsible for a fixed sector. This fixed mapping between the video id space to the lscache DNS namespace (logical video servers) makes it easier for individual YouTube front-end web servers (www.youtube.com) to generate independently and in a distributed fashion HTML pages with embedded URLs pointing to the relevant video(s) users are interested in, regardless where users are located or how logical servers are mapped to physical video servers or cache locations. Furthermore, there is also a fixed and consistent mapping between the (anycast) namespaces. For example, there is one-to-one mapping between the 192 DNS names of the lscache namespace and those of the tccache namespace, and a three-to-one maping between these namespaces and cache. These fixed mappings make it easy for each (physical) video server to decide given its logical name what portion of videos it is responsible for serving. DNS Resolution and (Coarse-grain) Locality-Aware Server Selection. The mapping between logical video servers (i.e., DNS names) and physical video servers (i.e., IP addresses) are done via DNS resolution. The mapping between DNS names to IP addresses are in general many-to-many: each

4 Table 1. Youtube Anycast and Unicast Namespaces. DNS namespace format # hostnames # IPs # prefixes # locations any/uni-cast lscache v[1-24].lscache[1-8].c.youtube.com 192 4, anycast tccache tc.v[1-24].cache[1-8].c.youtube.com anycast cache v[1-8].cache[1-8].c.youtube.com anycast rhost r[1-24].city[01-16][s,g,t][0-16].c.youtube.com 5, 044 5, unicast rhostisp r[1-24].isp-city[1-3].c.youtube.com unicast (anycast) DNS names are generally mapped to multiple IP addresses; and multiple DNS names may be mapped to the same IP address. YouTube employs a (coarse-grain) localityaware server selection strategy: depending on where the user request is originated, YouTube picks a primary video cache location that is close to the user, and resolves the requested lscache DNS name to one of the IP addresses within that location. Dynamic HTTP Request Re-direction. To perform finergrain and dynamic load-balancing, or to handle cache misses, YouTube employs HTTP request redirection mechanism. Such a redirection mechanism is especially useful and important, as YouTube always maps the user video request to a physical video server at a primary cache location. Since the size difference of the primary locations (in terms of the number of physical video servers or IP addresses) can be quite large, and the user demand is also likely to vary from one geographical area to another, dynamic load-balancing is needed. Further, the cache size at each location may also differ significantly, and videos cached at each location can change over time (e.g., due to the differing popularity of videos), cache misses are inevitable depending on how busy a video server at the primary location, it can either directly fetch a cold video from another video server at a secondary or tertiary location (e.g., via the Google internal backbone network), or re-direct the request directly to another video server at a secondary or tertiary location. YouTube cleverly utilizes the multiple (both anycast and unicast) DNS namespaces to perform dynamic load-balancing as well as to keep track of the redirection process. There is a strict ordering among the anycast namespaces, only redirection from a higher layer namespace to a lower layer namespace is allowed, e.g., from lscache to cache, but not the other way round; a redirection can jump across multiple layers, e.g., from lscache to tccache or cache. YouTube uses both a redirection count and a tag to keep track of the redirection sequences. 5. CACHE NAMESPACES & HIERARCHY In this section we describe the structure and organization of YouTube video caches. First, we describe several DNS namespaces used to refer to these cache servers, and how the mapping between video id space and the namespaces used by video cache servers is done. Finally, we describe how YouTube uses DNS infrastructure to direct the users to an IP address located in the vicinity of their location Anycast DNS Namespace Table 1 summarizes the anycast namespaces used by YouTube to refer to video cache servers. Our detailed analysis of video playback logs show that these anycast namespaces can be divided into following categories: Primary Video Caches. These are the hostnames embedded in the initial HTML file provided by the YouTube front-end web server to user, when a user accesses a video page. Using our analysis of initial HTML files for all the 434K collected from several vantage points, we found that there are a total of 192 such hostnames embedded in the main HTML file which refer to the servers hosting the Flash video objects. We also found that each video id maps to a unique lscache hostname out of 192 such names. For instance, a video identified using the video id MQCNuv2QxQY always maps to v23.lscache1.c.youtube.com lscache name from all the 1,314 vantage points at all times. Since all the video ids are uniformly distributed in the flat video identifier space, the number of video ids that map to each lscache hostname are also equally distributed. To demonstrate this we consider all the434k video ids and plot the number of video ids that map to each of the lscache hostnames in Figure 4. As seen in this figure, there are approximately equal number of videos mapped to each of the lscache hostnames. Based upon the location of IP addresses that hostnames in this namespace map to, we see 38 primary cache locations. We note that as we increased the number of vantage points, we increasingly uncovered additional primary cache locations. Secondary Video Caches. Similar to primary video caches, YouTube also deploys a secondary video cache for better availability and reliability. Again, it uses anycast DNS namespace to identify hosts in this set, which is of the following form: tc.v[1-24].cache[1-8].c.youtube.com. In addition, this namespace maps to a relatively smaller set of IP addresses in total, and as we will explain later, these video caches are located at only 8 locations. Tertiary Video Caches. Tertiary Video Caches show up as the final set of video caches during redirections in our playback logs. Unlike primary and secondary video caches, these tertiary video caches have a namespace constituted by only 64

5 Total views (as of Sep/2010) Videos x 10 5 Fig. 3. View counts for 434K videos Number of videos lscache hostnames Fig. 4. Number of videos mapped to each lscache hostname. Number of Planet Lab nodes that have at least one lscache hostnames mapped at the ith rank location Rank of the YouTube lscache Location Fig. 5. PlanetLab nodes with delay based rank of the YouTube location. DNS hostnames, which are represented using regular expression: v[1-8].cache[1-8].c.youtube.com. Since there are only 64 hostnames in the tertiary video cache namespace, hence there is a three-to-one mapping between the hostnames in primary/secondary video cache namespaces and tertiary video cache namespaces. Unlike primary cache locations, the number of secondary and tertiary cache locations did not increase as we added new vantage points Unicast DNS Namespace In addition to several anycast namespaces used by YouTube to refer to video cache servers, YouTube also uses a unicast namespace to identify individual video servers. These unicast DNS hostnames map to a unique IP address irrespective of the user location, it helps in redirecting the user to a specific server during the dynamic load balancing process Geo-mapping YouTube Cache Locations In order to geolocate YouTube IP addresses, first of all, we leverage the large number of unicast hostnames extracted using the video playback logs. As described earlier, each of these hostnames have 3-letter city codes embedded, which represent the nearest airport code for the corresponding YouTube location. We further verified using round trip delay measurements that these embedded city codes are correct. In the second step, we used the round trip delay logs for all the YouTube IP addresses collected using 471 PlanetLab nodes. We use a basic idea similar to the approach used by GeoPing [7]. In this approach, we consider the delay between an IP address and a set of PlanetLab nodes (vantage points) as the feature vector representing the IP address. Next, we cluster all these IP addresses using k-means clustering algorithm, and use Euclidean distance between the feature vectors as a distance measure. We assign each cluster a location, if we have at least one IP address in that cluster for which the location was already known using unicast hostnames. Also, in several cases we found that there were multiple such IP addresses in the clusters, for which the location was already Table 2. DNS resolutions for v1.lscache1.c.youtube.com PlanetLab node Resolved IP IP location adam.ee.ntu.edu.tw Taipei chimay.infonet.fundp.ac.be Amsterdam cs-planetlab4.cs.surrey.sfu.ca Seattle dannan.disy.inf.uni-konstanz.de Frankfurt ds-pl3.technion.ac.il Frankfurt known. In all such instances the location for these multiple IP addresses were always the same. In the end, we are able to geolocate all the YouTube IP addresses to 47 different cities spread across the globe. Furthermore, we found that primary caches are distributed in38 locations, secondary caches in 8 and tertiary caches in 5 locations with some locations hosting overlapping cache hierarchy. We plot these extracted YouTube locations on a world map in Figure Locality Aware DNS Resolution YouTube uses DNS-based location awareness to direct users to a nearby cache locations. As an example, in Table 2, we show DNS resolutions for v1.lscache1.c.youtube.com hostname performed from 5 different PlanetLab nodes. As seen in this table, based upon the location of the PlanetLab node, these hostnames mapped to an IP address at a nearby YouTube location. In order to verify location-awareness in DNS resolutions, we performed the following analysis. For each PlanetLab node, we order all 47 YouTube locations in the increasing order of round trip network delay and assign each YouTube location a rank in this order. Next, we consider lscache hostname to IP addresses mapping for each of the PlanetLab nodes, and see how they are distributed with respect to the rank of the corresponding YouTube location for the given PlanetLab node. In Figure 5 we plot the number of PlanetLab nodes which had at least one of lscache hostnames mapped to an ith rank YouTube location. As seen in this figure, more than 150 PlanetLab nodes have at least one of the IP addresses at the closest YouTube location with respect

6 to network delay. Only a very small number of nodes see that the mapped IP addresses are from farther locations. 6. HTTP REDIRECTIONS YouTube uses HTTP based redirection to achieve dynamic load-balancing. For instance, if the lscache server responsible for a video cannot serve the requested video, it sends a HTTP 302 response back to the client. This response tells the client to go to another server to download the video from. If the host corresponding to the HTTP redirect URL, again, can not provide the video for some reason, it sends another HTTP 302 response to the client to ask it to try yet another hostname. This redirection mechanism allows YouTube to perform dynamic load-sharing among its geographically distributed physical resources. In the following, we discuss the specific mechanisms used in these HTTP based redirections. To better understand the patterns in HTTP redirections, we carefully examined the video playback logs corresponding to such cases. Our analysis of these logs reveals several interesting patterns in these redirections and the mechanisms used to avoid HTTP redirection loops. The first key finding here is that HTTP redirections for any given video follow a very specific namespace hierarchy. A lscache video cache server may re-direct a video request to a corresponding tccache video cache server, or directly to a corresponding cache video cache server. Likewise, a tccache video cache server may only re-direct a video request to a corresponding cache video cache server but not to a lscache video server. In addition, we found that these HTTP redirections may also involve several unicast hostnames (such as rhost and rhostisp) as well. However, the redirection to one of these hostnames only occurs from one of the anycast hostnames. In the event when an anycast hostname redirects the user to one of the unicast hostnames, then it also updates the redirection URL with a tag corresponding to the tier of the forwarding anycast hostname. E.g., whenever an lscache host forwards the video request to rhost host, it adds a unique tag &st=lc in the redirection URL, which represents that request was forwarded from one of the host in lscache hostnames (tccache and cache use the tags &st=tcts and &st=ts respectively). These mechanisms ensures that the HTTP redirection always follow the cache hierarchy and does not create a redirection loop. As an ultimate protection against the possible redirection loops, redirection URL also contain a redirection counter, which is incremented by one every time a video request is forwarded from one anycast namespace to another anycast namespace. When the redirection counter reaches 4 and the corresponding anycast YouTube host can not serve the video, then it simply fails with a HTTP 503 error code. To confirm that maximum value for the redirection counter is 4, we modified the redirection URLs to have a redirection counter greater than 4, and asked several hosts to serve the video. In all such cases, we found that either the host served the video, or sent the HTTP 503 error code. 7. CONCLUSIONS AND FUTURE WORK In this paper, we built a globally distributed active measurement platform to reverse engineer YouTube video delivery framework and uncovered and geo-located YouTube s 3-tier physical video server hierarchy, and deduced the key design features of the YouTube video delivery cloud. We are confident that we have uncovered the major design features of the YouTube video delivery cloud. Nonetheless, there are still specific questions, such as the precise YouTube redirection decision logic and process, and how they are affected by video popularity, that still require in-depth analysis and additional experiments. Our current ongoing work, in part, attempts to answer these questions. While Google s YouTube video delivery cloud represents an example of the best practices in the design of such planet-scale systems, its design also poses several interesting and important questions regarding alternative architectural designs, cache placement, content replication and load balancing strategies, especially in terms of user perceived performance. In addition, the YouTube video delivery cloud design is clearly confined and constrained by the existing Internet architecture. Understanding the pros and cons in the YouTube video delivery cloud design also provides valuable insights into the future Internet architecture designs. 8. REFERENCES [1] Google to acquire youtube for $1.65 billion in stock. pressrel/google_youtube.html. [2] Google-YouTube: Bad News for Limelight? datacenterknowledge.com/archives/2006/10/ 13/google-youtube-bad-news-for-limelight/. [3] V. K. Adhikari, S. Jain, and Z. Zhang. YouTube Traffic Dynamics and Its Interplay with a Tier-1 ISP: An ISP Perspective. In IMC 10. ACM, [4] M. Cha, H. Kwak, P. Rodriguez, Y.-Y. Ahn, and S. Moon. I tube, you tube, everybody tubes: analyzing the world s largest user generated content video system. In IMC 07. ACM, [5] P. Gill, M. Arlitt, Z. Li, and A. Mahanti. Youtube traffic characterization: a view from the edge. In IMC 07. ACM, [6] C. Labovitz, S. Iekel-Johnson, D. McPherson, J. Oberheide, and F. Jahanian. Internet inter-domain traffic. In SIGCOMM 10. [7] V. N. Padmanabhan and L. Subramanian. An investigation of geographic mapping techniques for internet hosts. In SIGCOMM 01. [8] R. Zhou, S. Khemmarat, and L. Gao. The Impact of YouTube Recommendation System on Video Views. In IMC 10. [9] M. Zink, K. Suh, Y. Gu, and J. Kurose. Characteristics of youtube network traffic at a campus network - measurements, models, and implications. Comput. Netw., 2009.

How Do You Tube? Reverse Engineering the YouTube Video Delivery Cloud

How Do You Tube? Reverse Engineering the YouTube Video Delivery Cloud How Do You Tube? Reverse Engineering the YouTube Video Delivery Cloud Vijay Kumar Adhikari, Sourabh Jain, Yingying Chen, and Zhi Li Zhang Department of Computer Science & Engineering, University of Minnesota

More information

Where Do You Tube? Uncovering YouTube Server Selection Strategy

Where Do You Tube? Uncovering YouTube Server Selection Strategy Where Do You Tube? Uncovering YouTube Server Selection Strategy Vijay Kumar Adhikari, Sourabh Jain, Zhi-Li Zhang University of Minnesota- Twin Cities Abstract YouTube is one of the most popular video sharing

More information

A Tale of Three CDNs: An Active Measurement Study of Hulu and Its CDNs

A Tale of Three CDNs: An Active Measurement Study of Hulu and Its CDNs A Tale of Three CDNs: An Active Measurement Study of Hulu and Its CDNs Vijay Kumar Adhikari, Yang Guo, Fang Hao, Volker Hilt and Zhi-Li Zhang University of Minnesota, Bell-Labs/Alcatel-Lucent Abstract

More information

The Impact of YouTube Recommendation System on Video Views

The Impact of YouTube Recommendation System on Video Views The Impact of YouTube Recommendation System on Video Views Renjie Zhou, Samamon Khemmarat, Lixin Gao College of Computer Science and Technology Department of Electrical and Computer Engineering Harbin

More information

Internet Traffic Measurement

Internet Traffic Measurement Internet Traffic Measurement Internet Traffic Measurement Network Monitor Placement Measurement Analysis Tools Measurement Result Reporting Probing Mechanism Vantage Points Edge vs Core Hardware vs Software

More information

Experimentation with the YouTube Content Delivery Network (CDN)

Experimentation with the YouTube Content Delivery Network (CDN) Experimentation with the YouTube Content Delivery Network (CDN) Siddharth Rao Department of Computer Science Aalto University, Finland siddharth.rao@aalto.fi Sami Karvonen Department of Computer Science

More information

Politecnico di Torino. Porto Institutional Repository

Politecnico di Torino. Porto Institutional Repository Politecnico di Torino Porto Institutional Repository [Proceeding] Dissecting Video Server Selection Strategies in the YouTube CDN Original Citation: Ruben Torres,Alessandro Finamore,Jin Ryong Kim,Marco

More information

Overview. 15-441 15-441 Computer Networking 15-641. Lecture 18: Delivering Content Content Delivery Networks Peter Steenkiste

Overview. 15-441 15-441 Computer Networking 15-641. Lecture 18: Delivering Content Content Delivery Networks Peter Steenkiste Overview 5-44 5-44 Computer Networking 5-64 Lecture 8: Delivering Content Content Delivery Networks Peter Steenkiste Fall 04 www.cs.cmu.edu/~prs/5-44-f4 Web Consistent hashing Peer-to-peer CDN Motivation

More information

YouTube Traffic Dynamics and Its Interplay with a Tier-1 ISP: An ISP Perspective

YouTube Traffic Dynamics and Its Interplay with a Tier-1 ISP: An ISP Perspective YouTube Traffic Dynamics and Its Interplay with a Tier- ISP: An ISP Perspective Vijay Kumar Adhikari, Sourabh Jain, and Zhi-Li Zhang Dept. of Computer Science, University of Minnesota Minneapolis, MN viadhi@cs.umn.edu,

More information

Measuring the Web: Part I - - Content Delivery Networks. Prof. Anja Feldmann, Ph.D. Dr. Ramin Khalili Georgios Smaragdakis, PhD

Measuring the Web: Part I - - Content Delivery Networks. Prof. Anja Feldmann, Ph.D. Dr. Ramin Khalili Georgios Smaragdakis, PhD Measuring the Web: Part I - - Content Delivery Networks Prof. Anja Feldmann, Ph.D. Dr. Ramin Khalili Georgios Smaragdakis, PhD Acknowledgement Material presented in these slides is borrowed from presentajons

More information

Web Application Hosting Cloud Architecture

Web Application Hosting Cloud Architecture Web Application Hosting Cloud Architecture Executive Overview This paper describes vendor neutral best practices for hosting web applications using cloud computing. The architectural elements described

More information

Indirection. science can be solved by adding another level of indirection" -- Butler Lampson. "Every problem in computer

Indirection. science can be solved by adding another level of indirection -- Butler Lampson. Every problem in computer Indirection Indirection: rather than reference an entity directly, reference it ( indirectly ) via another entity, which in turn can or will access the original entity A x B "Every problem in computer

More information

Content Delivery Networks (CDN) Dr. Yingwu Zhu

Content Delivery Networks (CDN) Dr. Yingwu Zhu Content Delivery Networks (CDN) Dr. Yingwu Zhu Web Cache Architecure Local ISP cache cdn Reverse Reverse Proxy Reverse Proxy Reverse Proxy Proxy L4 Switch Content Content Content Server Content Server

More information

Increasing the Coverage of Vantage Points in Distributed Active Network Measurements by Crowdsourcing

Increasing the Coverage of Vantage Points in Distributed Active Network Measurements by Crowdsourcing NOTICE: This is the author s version of a work accepted for publication by Springer. Changes resulting from the publishing process, including editing, corrections, structural formatting and other quality

More information

Content Delivery and the Natural Evolution of DNS

Content Delivery and the Natural Evolution of DNS Content Delivery and the Natural Evolution of DNS Remote DNS Trends, Performance Issues and Alternative Solutions John S. Otto Mario A. Sánchez John P. Rula Fabián E. Bustamante Northwestern University

More information

Measuring YouTube Quality of Experience for Users in Residential ISPs*

Measuring YouTube Quality of Experience for Users in Residential ISPs* Measuring YouTube Quality of Experience for Users in Residential ISPs* Deep Medhi dmedhi@umkc.edu University of Missouri Kansas City February 2013 *Joint work with Parikshit Juluri (UMKC), Louis Plissonneau

More information

DATA COMMUNICATOIN NETWORKING

DATA COMMUNICATOIN NETWORKING DATA COMMUNICATOIN NETWORKING Instructor: Ouldooz Baghban Karimi Course Book: Computer Networking, A Top-Down Approach, Kurose, Ross Slides: - Course book Slides - Slides from Princeton University COS461

More information

Web Caching and CDNs. Aditya Akella

Web Caching and CDNs. Aditya Akella Web Caching and CDNs Aditya Akella 1 Where can bottlenecks occur? First mile: client to its ISPs Last mile: server to its ISP Server: compute/memory limitations ISP interconnections/peerings: congestion

More information

Analyzing Caching Benefits for YouTube Traffic in Edge Networks - A Measurement-Based Evaluation

Analyzing Caching Benefits for YouTube Traffic in Edge Networks - A Measurement-Based Evaluation Analyzing Caching Benefits for YouTube Traffic in Edge Networks - A Measurement-Based Evaluation Lothar Braun, Alexander Klein, Georg Carle {braun,klein,carle}@net.in.tum.de Technische Universität München

More information

Analyzing the Impact of YouTube Delivery Policies on User Experience

Analyzing the Impact of YouTube Delivery Policies on User Experience Analyzing the Impact of YouTube Delivery Policies on User Experience Louis Plissonneau Orange Labs France louis.plissonneau@orange.com Ernst Biersack Eurecom France erbi@eurecom.fr Parikshit Juluri University

More information

Testing & Assuring Mobile End User Experience Before Production. Neotys

Testing & Assuring Mobile End User Experience Before Production. Neotys Testing & Assuring Mobile End User Experience Before Production Neotys Agenda Introduction The challenges Best practices NeoLoad mobile capabilities Mobile devices are used more and more At Home In 2014,

More information

Web Analytics Understand your web visitors without web logs or page tags and keep all your data inside your firewall.

Web Analytics Understand your web visitors without web logs or page tags and keep all your data inside your firewall. Web Analytics Understand your web visitors without web logs or page tags and keep all your data inside your firewall. 5401 Butler Street, Suite 200 Pittsburgh, PA 15201 +1 (412) 408 3167 www.metronomelabs.com

More information

Global Server Load Balancing

Global Server Load Balancing White Paper Overview Many enterprises attempt to scale Web and network capacity by deploying additional servers and increased infrastructure at a single location, but centralized architectures are subject

More information

Request Routing, Load-Balancing and Fault- Tolerance Solution - MediaDNS

Request Routing, Load-Balancing and Fault- Tolerance Solution - MediaDNS White paper Request Routing, Load-Balancing and Fault- Tolerance Solution - MediaDNS June 2001 Response in Global Environment Simply by connecting to the Internet, local businesses transform themselves

More information

Web Traffic Capture. 5401 Butler Street, Suite 200 Pittsburgh, PA 15201 +1 (412) 408 3167 www.metronomelabs.com

Web Traffic Capture. 5401 Butler Street, Suite 200 Pittsburgh, PA 15201 +1 (412) 408 3167 www.metronomelabs.com Web Traffic Capture Capture your web traffic, filtered and transformed, ready for your applications without web logs or page tags and keep all your data inside your firewall. 5401 Butler Street, Suite

More information

On the Feasibility of Prefetching and Caching for Online TV Services: A Measurement Study on Hulu

On the Feasibility of Prefetching and Caching for Online TV Services: A Measurement Study on Hulu On the Feasibility of Prefetching and Caching for Online TV Services: A Measurement Study on Hulu Dilip Kumar Krishnappa, Samamon Khemmarat, Lixin Gao, Michael Zink University of Massachusetts Amherst,

More information

GLOBAL SERVER LOAD BALANCING WITH SERVERIRON

GLOBAL SERVER LOAD BALANCING WITH SERVERIRON APPLICATION NOTE GLOBAL SERVER LOAD BALANCING WITH SERVERIRON Growing Global Simply by connecting to the Internet, local businesses transform themselves into global ebusiness enterprises that span the

More information

Analysing the impact of CDN based service delivery on traffic engineering

Analysing the impact of CDN based service delivery on traffic engineering Analysing the impact of CDN based service delivery on traffic engineering Gerd Windisch Chair for Communication Networks Technische Universität Chemnitz Gerd Windisch - Chair for Communication Networks

More information

A DNS Reflection Method for Global Traffic Management

A DNS Reflection Method for Global Traffic Management A DNS Reflection Method for Global Traffic Management Cheng Huang Microsoft Research Albert Greenberg Microsoft Research Nick Holt Microsoft Corporation Jin Li Microsoft Research Y. Angela Wang Polytechnic

More information

Craig Labovitz, Scott Iekel-Johnson, Danny McPherson Arbor Networks Jon Oberheide, Farnam Jahanian University of Michigan

Craig Labovitz, Scott Iekel-Johnson, Danny McPherson Arbor Networks Jon Oberheide, Farnam Jahanian University of Michigan Internet Inter-Domain Traffic Craig Labovitz, Scott Iekel-Johnson, Danny McPherson Arbor Networks Jon Oberheide, Farnam Jahanian University of Michigan Motivation Measuring the Internet is hard Significant

More information

LARGE-SCALE INTERNET MEASUREMENTS FOR DIAGNOSTICS AND PUBLIC POLICY. Henning Schulzrinne (+ Walter Johnston & James Miller) FCC & Columbia University

LARGE-SCALE INTERNET MEASUREMENTS FOR DIAGNOSTICS AND PUBLIC POLICY. Henning Schulzrinne (+ Walter Johnston & James Miller) FCC & Columbia University 1 LARGE-SCALE INTERNET MEASUREMENTS FOR DIAGNOSTICS AND PUBLIC POLICY Henning Schulzrinne (+ Walter Johnston & James Miller) FCC & Columbia University 2 Overview Quick overview What does MBA measure? Can

More information

Internet Content Distribution

Internet Content Distribution Internet Content Distribution Chapter 4: Content Distribution Networks (TUD Student Use Only) Chapter Outline Basics of content distribution networks (CDN) Why CDN? How do they work? Client redirection

More information

DISSECTING VIDEO SERVER SELECTION STRATEGIES IN THE CDN [ICDCS 2011]

DISSECTING VIDEO SERVER SELECTION STRATEGIES IN THE CDN [ICDCS 2011] DISSECTING VIDEO SERVER SELECTION STRATEGIES IN THE CDN [ICDCS 2011] Alessandro Finamore Marco Mellia Maurizio Munafò Ruben Torres Sanjay Rao 2nd TMA PhD School Objectives 1 YouTube is the most popular

More information

Week 3 / Paper 2. Bernhard Ager, Wolfgang Mühlbauer, Georgios Smaragdakis, Steve Uhlig ACM IMC 2010.

Week 3 / Paper 2. Bernhard Ager, Wolfgang Mühlbauer, Georgios Smaragdakis, Steve Uhlig ACM IMC 2010. Week 3 / Paper 2 Comparing DNS Resolvers in the Wild Bernhard Ager, Wolfgang Mühlbauer, Georgios Smaragdakis, Steve Uhlig ACM IMC 2010. Main point How does ISP DNS compare with Google DNS and OpenDNS?

More information

Exploring YouTube s Content Distribution Network Through Distributed Application-Layer Measurements: A First View

Exploring YouTube s Content Distribution Network Through Distributed Application-Layer Measurements: A First View Exploring YouTube s Content Distribution Network Through Distributed Application-Layer Measurements: A First View Albert Rafetseder, Florian Metzger, David Stezenbach, and Kurt Tutschku {albert.rafetseder,

More information

1. Comments on reviews a. Need to avoid just summarizing web page asks you for:

1. Comments on reviews a. Need to avoid just summarizing web page asks you for: 1. Comments on reviews a. Need to avoid just summarizing web page asks you for: i. A one or two sentence summary of the paper ii. A description of the problem they were trying to solve iii. A summary of

More information

From Internet Data Centers to Data Centers in the Cloud

From Internet Data Centers to Data Centers in the Cloud From Internet Data Centers to Data Centers in the Cloud This case study is a short extract from a keynote address given to the Doctoral Symposium at Middleware 2009 by Lucy Cherkasova of HP Research Labs

More information

CDN and Traffic-structure

CDN and Traffic-structure CDN and Traffic-structure Outline Basics CDN Traffic Analysis 2 Outline Basics CDN Building Blocks Services Evolution Traffic Analysis 3 A Centralized Web! Slow content must traverse multiple backbones

More information

A Survey on Content Delivery of Web-Pages

A Survey on Content Delivery of Web-Pages International Journal of Computer Sciences and Engineering Open Access Survey Paper Volume-4, Issue-3 E-ISSN: 2347-2693 A Survey on Content Delivery of Web-Pages Aaqib Bashir 1 and Prof. T. H. Gurav 2

More information

Lecture 3: Scaling by Load Balancing 1. Comments on reviews i. 2. Topic 1: Scalability a. QUESTION: What are problems? i. These papers look at

Lecture 3: Scaling by Load Balancing 1. Comments on reviews i. 2. Topic 1: Scalability a. QUESTION: What are problems? i. These papers look at Lecture 3: Scaling by Load Balancing 1. Comments on reviews i. 2. Topic 1: Scalability a. QUESTION: What are problems? i. These papers look at distributing load b. QUESTION: What is the context? i. How

More information

Internet Traffic Evolution 2007-2011

Internet Traffic Evolution 2007-2011 Internet Traffic Evolution 2007-2011 Craig Labovitz April 6, 2011 Talk Outline Four-year ongoing inter-domain traffic study Review of 2010 results (NANOG / IETF / SIGCOMM) Methodology Changing carrier

More information

Handling Flash Crowds From Your Garage

Handling Flash Crowds From Your Garage Handling Flash Crowds From Your Garage Jeremy Elson and Jon Howell Microsoft Research USENIX ATC 2008 Scaling For Many Users Scaling For Many Users... quickly Scaling For Many Users... quickly... on a

More information

Content Distribu-on Networks (CDNs)

Content Distribu-on Networks (CDNs) Content Distribu-on Networks (CDNs) Jennifer Rexford COS 461: Computer Networks Lectures: MW 10-10:0am in Architecture N101 hjp://www.cs.princeton.edu/courses/archive/spr12/cos461/ Second Half of the Course

More information

ThreatSpike Dome: A New Approach To Security Monitoring

ThreatSpike Dome: A New Approach To Security Monitoring ThreatSpike Dome: A New Approach To Security Monitoring 2015 ThreatSpike Labs Limited The problem with SIEM Hacking, insider and advanced persistent threats can be difficult to detect with existing product

More information

Internet Traffic and Content Consolidation

Internet Traffic and Content Consolidation Internet Traffic and Content Consolidation Craig Labovitz Chief Scientist, Arbor Networks S. Iekel-Johnson, D. McPherson Arbor Networks, Inc. J. Oberheide, F. Jahanian University of Michigan Talk Outline

More information

Concept of Cache in web proxies

Concept of Cache in web proxies Concept of Cache in web proxies Chan Kit Wai and Somasundaram Meiyappan 1. Introduction Caching is an effective performance enhancing technique that has been used in computer systems for decades. However,

More information

Analysis of YouTube User Experience from Passive Measurements

Analysis of YouTube User Experience from Passive Measurements Analysis of YouTube User Experience from Passive Measurements Giorgos Dimopoulos Pere Barlet-Ros Josep Sanjuàs-Cuxart Department of Computer Architecture UPC, Barcelonatech {gd, pbarlet, jsanjuas}@ac.upc.edu

More information

A Study of YouTube recommendation graph based on measurements and stochastic tools

A Study of YouTube recommendation graph based on measurements and stochastic tools A Study of YouTube recommendation graph based on measurements and stochastic tools Yonathan Portilla, Alexandre Reiffers, Eitan Altman, Rachid El-Azouzi ABSTRACT The Youtube recommendation is one the most

More information

Load Balancing. Final Network Exam LSNAT. Sommaire. How works a "traditional" NAT? Un article de Le wiki des TPs RSM.

Load Balancing. Final Network Exam LSNAT. Sommaire. How works a traditional NAT? Un article de Le wiki des TPs RSM. Load Balancing Un article de Le wiki des TPs RSM. PC Final Network Exam Sommaire 1 LSNAT 1.1 Deployement of LSNAT in a globally unique address space (LS-NAT) 1.2 Operation of LSNAT in conjunction with

More information

A First Look at Inter-Data Center Traffic Characteristics via Yahoo! Datasets

A First Look at Inter-Data Center Traffic Characteristics via Yahoo! Datasets A First Look at Inter-Data Center Traffic Characteristics via Yahoo! Datasets Yingying Chen, Sourabh Jain, Vijay Kumar Adhikari, Zhi-Li Zhang, and Kuai Xu 2 University of Minnesota-Twin Cities 2 Arizona

More information

Peer-to-Peer Networks. Chapter 6: P2P Content Distribution

Peer-to-Peer Networks. Chapter 6: P2P Content Distribution Peer-to-Peer Networks Chapter 6: P2P Content Distribution Chapter Outline Content distribution overview Why P2P content distribution? Network coding Peer-to-peer multicast Kangasharju: Peer-to-Peer Networks

More information

HOW TO CONFIGURE PASS-THRU PROXY FOR ORACLE APPLICATIONS

HOW TO CONFIGURE PASS-THRU PROXY FOR ORACLE APPLICATIONS HOW TO CONFIGURE PASS-THRU PROXY FOR ORACLE APPLICATIONS Overview of Oracle JInitiator Oracle JInitiator enables users to run Oracle Forms applications using Netscape Navigator or Internet Explorer. It

More information

FortiBalancer: Global Server Load Balancing WHITE PAPER

FortiBalancer: Global Server Load Balancing WHITE PAPER FortiBalancer: Global Server Load Balancing WHITE PAPER FORTINET FortiBalancer: Global Server Load Balancing PAGE 2 Introduction Scalability, high availability and performance are critical to the success

More information

Web Browsing Quality of Experience Score

Web Browsing Quality of Experience Score Web Browsing Quality of Experience Score A Sandvine Technology Showcase Contents Executive Summary... 1 Introduction to Web QoE... 2 Sandvine s Web Browsing QoE Metric... 3 Maintaining a Web Page Library...

More information

Getting Started with AWS. Hosting a Static Website

Getting Started with AWS. Hosting a Static Website Getting Started with AWS Hosting a Static Website Getting Started with AWS: Hosting a Static Website Copyright 2016 Amazon Web Services, Inc. and/or its affiliates. All rights reserved. Amazon's trademarks

More information

Using NCache for ASP.NET Sessions in Web Farms

Using NCache for ASP.NET Sessions in Web Farms Using NCache for ASP.NET Sessions in Web Farms April 22, 2015 Contents 1 Getting Started... 1 1.1 Step 1: Install NCache... 1 1.2 Step 2: Configure for Multiple Network Cards... 1 1.3 Step 3: Configure

More information

Measuring and Evaluating Large-Scale CDNs

Measuring and Evaluating Large-Scale CDNs Measuring and Evaluating Large-Scale CDNs Cheng Huang Microsoft Research Redmond, WA 98052 Angela Wang Polytechnic Institute of NYU Brooklyn, NY 11201 Jin Li Microsoft Research Redmond, WA 98052 Keith

More information

International Journal of Advanced Research in Computer Science and Software Engineering

International Journal of Advanced Research in Computer Science and Software Engineering Volume 2, Issue 9, September 2012 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com An Experimental

More information

A First Look at Inter-Data Center Traffic Characteristics via Yahoo! Datasets

A First Look at Inter-Data Center Traffic Characteristics via Yahoo! Datasets A First Look at Inter-Data Center Traffic Characteristics via Yahoo! Datasets Yingying Chen, Sourabh Jain,VijayKumarAdhikari, Zhi-Li Zhang,andKuaiXu 2 University of Minnesota-Twin Cities 2 Arizona State

More information

Cisco Videoscape Distribution Suite Service Broker

Cisco Videoscape Distribution Suite Service Broker Data Sheet Cisco Videoscape Distribution Suite Service Broker Product Overview Cisco Videoscape Distribution Suite Service Broker (VDS-SB) encompasses a broad range of capabilities, particularly in accelerating

More information

Unreeling Netflix: Understanding and Improving Multi-CDN Movie Delivery

Unreeling Netflix: Understanding and Improving Multi-CDN Movie Delivery Unreeling Netflix: Understanding and Improving Multi-CDN Movie Delivery Vijay Kumar Adhikari, Yang Guo, Fang Hao, Matteo Varvello, Volker Hilt, Moritz Steiner and Zhi-Li Zhang University of Minnesota,

More information

ATLAS Internet Observatory Bandwidth Bandwagon: An ISOC Briefing Panel November 11, 2009, Hiroshima, Japan

ATLAS Internet Observatory Bandwidth Bandwagon: An ISOC Briefing Panel November 11, 2009, Hiroshima, Japan ATLAS Internet Observatory Bandwidth Bandwagon: An ISOC Briefing Panel November 11, 2009, Hiroshima, Japan Danny McPherson danny@arbor.net Chief Security Officer ATLAS Observatory Details ISP / Content

More information

Information- Centric Networks. Section # 3.2: DNS Issues Instructor: George Xylomenos Department: Informatics

Information- Centric Networks. Section # 3.2: DNS Issues Instructor: George Xylomenos Department: Informatics Information- Centric Networks Section # 3.2: DNS Issues Instructor: George Xylomenos Department: Informatics Funding These educational materials have been developed as part of the instructors educational

More information

Accelerating Wordpress for Pagerank and Profit

Accelerating Wordpress for Pagerank and Profit Slide No. 1 Accelerating Wordpress for Pagerank and Profit Practical tips and tricks to increase the speed of your site, improve conversions and climb the search rankings By: Allan Jude November 2011 Vice

More information

Disaster Recovery Design Ehab Ashary University of Colorado at Colorado Springs

Disaster Recovery Design Ehab Ashary University of Colorado at Colorado Springs Disaster Recovery Design Ehab Ashary University of Colorado at Colorado Springs As a head of the campus network department in the Deanship of Information Technology at King Abdulaziz University for more

More information

D. SamKnows Methodology 20 Each deployed Whitebox performs the following tests: Primary measure(s)

D. SamKnows Methodology 20 Each deployed Whitebox performs the following tests: Primary measure(s) v. Test Node Selection Having a geographically diverse set of test nodes would be of little use if the Whiteboxes running the test did not have a suitable mechanism to determine which node was the best

More information

Combining Global Load Balancing and Geo-location with Emissary TM

Combining Global Load Balancing and Geo-location with Emissary TM Combining Global Load Balancing and Geo-location with Emissary TM A New Kind of Global Internet Traffic Management Appliance from Coyote Point Systems and Digital Envoy Establishing a Geo-Sensitive, Highly

More information

ABSTRACT Acknowledgments List of Abbreviations Contents ABSTRACT 3 Acknowledgments 5 List of Abbreviations 7 List of Figures 15 List of Tables 23 1 Introduction 25 2 Motivation and background 29 3 Overview

More information

Dynamics of Prefix Usage at an Edge Router

Dynamics of Prefix Usage at an Edge Router Dynamics of Prefix Usage at an Edge Router Kaustubh Gadkari, Daniel Massey, and Christos Papadopoulos Computer Science Department, Colorado State University, USA {kaustubh, massey, christos@cs.colostate.edu}

More information

Distributed Systems. 25. Content Delivery Networks (CDN) 2014 Paul Krzyzanowski. Rutgers University. Fall 2014

Distributed Systems. 25. Content Delivery Networks (CDN) 2014 Paul Krzyzanowski. Rutgers University. Fall 2014 Distributed Systems 25. Content Delivery Networks (CDN) Paul Krzyzanowski Rutgers University Fall 2014 November 16, 2014 2014 Paul Krzyzanowski 1 Motivation Serving web content from one location presents

More information

Object Popularity Distributions in Online Social Networks

Object Popularity Distributions in Online Social Networks Object Popularity Distributions in Online Social Networks Theo Lins Computer Science Dept. Federal University of Ouro Preto (UFOP) Ouro Preto, Brazil theosl@gmail.com Wellington Dores Computer Science

More information

How Akamai Maps the Net:

How Akamai Maps the Net: How Akamai Maps the Net: An Industry Perspective George Economou In 2010, everyone uses the Internet. Even if you don t browse the Web, your computer, DVD player, and other appliances try to pull uses

More information

Content Delivery Networks. Shaxun Chen April 21, 2009

Content Delivery Networks. Shaxun Chen April 21, 2009 Content Delivery Networks Shaxun Chen April 21, 2009 Outline Introduction to CDN An Industry Example: Akamai A Research Example: CDN over Mobile Networks Conclusion Outline Introduction to CDN An Industry

More information

Evaluation Guide. Powerful & Immediate Business Web Security via the Cloud

Evaluation Guide. Powerful & Immediate Business Web Security via the Cloud Evaluation Guide Powerful & Immediate Business Web Security via the Cloud Contents 1 Introduction & Product highlights 2 Set up & Configuration 3 Managing your WebTitan Cloud Service 4 Reporting 5 Support

More information

Simple Solution for a Location Service. Naming vs. Locating Entities. Forwarding Pointers (2) Forwarding Pointers (1)

Simple Solution for a Location Service. Naming vs. Locating Entities. Forwarding Pointers (2) Forwarding Pointers (1) Naming vs. Locating Entities Till now: resources with fixed locations (hierarchical, caching,...) Problem: some entity may change its location frequently Simple solution: record aliases for the new address

More information

Content Retrieval using Cloud-based DNS

Content Retrieval using Cloud-based DNS Content Retrieval using Cloud-based DNS Ravish Khosla, Sonia Fahmy, Y. Charlie Hu Purdue University Email: {rkhosla, fahmy, ychu}@purdue.edu Abstract Cloud-computing systems are rapidly gaining momentum,

More information

ANATOMY OF A DDoS ATTACK AGAINST THE DNS INFRASTRUCTURE

ANATOMY OF A DDoS ATTACK AGAINST THE DNS INFRASTRUCTURE ANATOMY OF A DDoS ATTACK AGAINST THE DNS INFRASTRUCTURE ANATOMY OF A DDOS ATTACK AGAINST THE DNS INFRASTRUCTURE The Domain Name System (DNS) is part of the functional infrastructure of the Internet and

More information

State of the Cloud DNS Report

State of the Cloud DNS Report transparency for the cloud State of the Cloud DNS Report Basic Edition April 2015 2015 Table of Contents Overview Introduction 3 Anycast vs. Unicast DNS 3 Provider Overview & Current News 4 Provider Marketshare

More information

Traffic, Caching & Traffic ACROSS Data Centers in Clouds, CDNs and ISP Networks

Traffic, Caching & Traffic ACROSS Data Centers in Clouds, CDNs and ISP Networks Traffic, in Clouds, CDNs and ISP Networks, Deutsche Telekom, Darmstadt, Germany E-Mail: gerhard.hasslinger@telekom.de Caching options, main goals and benefits in Ip networks CDN, cloud traffic in ISP networks

More information

The OpenDNS Global Network Delivers a Secure Connection Every Time. Everywhere.

The OpenDNS Global Network Delivers a Secure Connection Every Time. Everywhere. The OpenDNS Global Network Delivers a Secure Connection Every Time. Everywhere. Network Performance Users devices create multiple simultaneous connections each time we exchange data with other Internet

More information

Web Analytics Definitions Approved August 16, 2007

Web Analytics Definitions Approved August 16, 2007 Web Analytics Definitions Approved August 16, 2007 Web Analytics Association 2300 M Street, Suite 800 Washington DC 20037 standards@webanalyticsassociation.org 1-800-349-1070 Licensed under a Creative

More information

Computer Networks 53 (2009) 501 514. Contents lists available at ScienceDirect. Computer Networks. journal homepage: www.elsevier.

Computer Networks 53 (2009) 501 514. Contents lists available at ScienceDirect. Computer Networks. journal homepage: www.elsevier. Computer Networks 53 (2009) 501 514 Contents lists available at ScienceDirect Computer Networks journal homepage: www.elsevier.com/locate/comnet Characteristics of YouTube network traffic at a campus network

More information

BM 465E Distributed Systems

BM 465E Distributed Systems BM 465E Distributed Systems Lecture 4 Networking (cont.) Mehmet Demirci Today Overlay networks Data centers Content delivery networks Overlay Network A virtual network built on top of another network Overlay

More information

Trends in Internet Traffic Patterns Darren Anstee, EMEA Solutions Architect

Trends in Internet Traffic Patterns Darren Anstee, EMEA Solutions Architect Trends in Internet Traffic Patterns Darren Anstee, EMEA Solutions Architect This Talk The End of the Internet as we Know it We present the largest study of Internet traffic every conducted Peer-reviewed

More information

DOCUMENT REFERENCE: SQ309-002-EN. SAMKNOWS TEST METHODOLOGY Web-based Broadband Performance White Paper. July 2015

DOCUMENT REFERENCE: SQ309-002-EN. SAMKNOWS TEST METHODOLOGY Web-based Broadband Performance White Paper. July 2015 DOCUMENT REFERENCE: SQ309-002-EN SAMKNOWS TEST METHODOLOGY Web-based Broadband Performance White Paper July 2015 SAMKNOWS QUALITY CONTROLLED DOCUMENT. SQ REV LANG STATUS OWNER DATED 309 03 EN FINAL SC

More information

Design and Evaluation of a Real-Time URL Spam Filtering Service

Design and Evaluation of a Real-Time URL Spam Filtering Service Design and Evaluation of a Real-Time URL Spam Filtering Service Geraldo Franciscani 15 de Maio de 2012 Teacher: Ponnurangam K (PK) Introduction Initial Presentation Monarch is a real-time system for filtering

More information

Inside Dropbox: Understanding Personal Cloud Storage Services

Inside Dropbox: Understanding Personal Cloud Storage Services Inside Dropbox: Understanding Personal Cloud Storage Services Idilio Drago Marco Mellia Maurizio M. Munafò Anna Sperotto Ramin Sadre Aiko Pras IMC 2012 Boston Motivation and goals 1/14 Personal cloud storage

More information

The Effect of Caches for Mobile Broadband Internet Access

The Effect of Caches for Mobile Broadband Internet Access The Effect of s for Mobile Jochen Eisl, Nokia Siemens Networks, Munich, Germany Haßlinger, Deutsche Telekom Technik,, Darmstadt, Germany IP-based content delivery: CDN & cache architecture Impact of access

More information

ATLAS Internet Observatory 2009 Annual Report

ATLAS Internet Observatory 2009 Annual Report ATLAS Internet Observatory 2009 Annual Report C. Labovitz, S. Iekel-Johnson, D. McPherson Arbor Networks, Inc. J. Oberheide, F. Jahanian University of Michigan M. Karir Merit Network, Inc. ATLAS Internet

More information

Research of User Experience Centric Network in MBB Era ---Service Waiting Time

Research of User Experience Centric Network in MBB Era ---Service Waiting Time Research of User Experience Centric Network in MBB ---Service Waiting Time Content Content 1 Introduction... 1 1.1 Key Findings... 1 1.2 Research Methodology... 2 2 Explorer Excellent MBB Network Experience

More information

Monitoring Pramati Web Server

Monitoring Pramati Web Server Monitoring Pramati Web Server 15 Overview This section describes how to monitor Pramati Web Server from the Console. You can monitor information regarding the running Default Server and Virtual Hosts,

More information

International Journal of Scientific & Engineering Research, Volume 4, Issue 11, November-2013 349 ISSN 2229-5518

International Journal of Scientific & Engineering Research, Volume 4, Issue 11, November-2013 349 ISSN 2229-5518 International Journal of Scientific & Engineering Research, Volume 4, Issue 11, November-2013 349 Load Balancing Heterogeneous Request in DHT-based P2P Systems Mrs. Yogita A. Dalvi Dr. R. Shankar Mr. Atesh

More information

Internet Firewall CSIS 4222. Packet Filtering. Internet Firewall. Examples. Spring 2011 CSIS 4222. net15 1. Routers can implement packet filtering

Internet Firewall CSIS 4222. Packet Filtering. Internet Firewall. Examples. Spring 2011 CSIS 4222. net15 1. Routers can implement packet filtering Internet Firewall CSIS 4222 A combination of hardware and software that isolates an organization s internal network from the Internet at large Ch 27: Internet Routing Ch 30: Packet filtering & firewalls

More information

Distributed Systems. 24. Content Delivery Networks (CDN) 2013 Paul Krzyzanowski. Rutgers University. Fall 2013

Distributed Systems. 24. Content Delivery Networks (CDN) 2013 Paul Krzyzanowski. Rutgers University. Fall 2013 Distributed Systems 24. Content Delivery Networks (CDN) Paul Krzyzanowski Rutgers University Fall 2013 November 27, 2013 2013 Paul Krzyzanowski 1 Motivation Serving web content from one location presents

More information

ProxySG TechBrief Implementing a Reverse Proxy

ProxySG TechBrief Implementing a Reverse Proxy ProxySG TechBrief Implementing a Reverse Proxy What is a reverse proxy? The Blue Coat ProxySG provides the basis for a robust and flexible Web communications solution. In addition to Web policy management,

More information

Meeting Worldwide Demand for your Content

Meeting Worldwide Demand for your Content Meeting Worldwide Demand for your Content Evolving to a Content Delivery Network A Lucent Technologies White Paper By L. R. Beaumont 4/25/01 Meeting Worldwide Demand for your Content White Paper Table

More information

HOW IS WEB APPLICATION DEVELOPMENT AND DELIVERY CHANGING?

HOW IS WEB APPLICATION DEVELOPMENT AND DELIVERY CHANGING? WHITE PAPER : WEB PERFORMANCE TESTING Why Load Test at all? The reason we load test is to ensure that people using your web site can successfully access the pages and complete whatever kind of transaction

More information

Front-End Performance Testing and Optimization

Front-End Performance Testing and Optimization Front-End Performance Testing and Optimization Abstract Today, web user turnaround starts from more than 3 seconds of response time. This demands performance optimization on all application levels. Client

More information

Internet Video Delivery in YouTube: From Traffic Measurements to Quality of Experience

Internet Video Delivery in YouTube: From Traffic Measurements to Quality of Experience Internet Video Delivery in YouTube: From Traffic Measurements to Quality of Experience Tobias Hoßfeld 1, Raimund Schatz 2, Ernst Biersack 3, and Louis Plissonneau 4 1 University of Würzburg, Institute

More information

Chapter-1 : Introduction 1 CHAPTER - 1. Introduction

Chapter-1 : Introduction 1 CHAPTER - 1. Introduction Chapter-1 : Introduction 1 CHAPTER - 1 Introduction This thesis presents design of a new Model of the Meta-Search Engine for getting optimized search results. The focus is on new dimension of internet

More information